Comments on: Justification and Action http://jonathanichikawa.net/weblog/justification-and-action/ Jonathan Ichikawa's website Thu, 29 Jul 2010 13:07:38 +0000 http://wordpress.org/?v=2.7.1 hourly 1 By: Matt McGrath http://jonathanichikawa.net/weblog/justification-and-action/#comment-288 Matt McGrath Fri, 12 Feb 2010 15:46:49 +0000 http://jonathanichikawa.net/?p=189#comment-288 Ok, so we do agree that, given JJ, fallibilism, and moderate externalism, we get the result Fantl and I mentioned (i.e., that there can be internal duplicates, with the same options available, who differ in what they are justified in doing). I outlined how we get this result in my comment #3 -- but, I agree, that to fallibilism has to be in there. We could talk about that. I take that to be an assumption that a lot of externalists would want to make (you don't need epistemic probability 1 = justified certainty) to have justified belief. The more I think about it, though, I wonder if we even *need* to bring in JJ to get a problem for moderate externalism. Consider moderate externalism about epistemic probability (or degree of justification if you like). This would hold that how probable p was for you was a function of externalistic factors. Now, take someone for whom p has a .7 probability, and suppose given her stakes, the option of doing A has an higher expected value than the option of doing B (and suppose those are the only choices). Call this person S1. It seems to me it should be possible for there to be a subject S2 with the same stakes and available options and who is internally indistinguishable from S1 but whose probability is very different (say .1), because of differences in the reliability of the available processes, or degree of safety of the various indications available, etc. We can imagine that for S2 the option of doing B has a higher expected value than that of doing A because of this difference in the probability of p for them. Then S1 and S2 would be alike internally, have the same options available, but differ in what they are justified in doing. You can illustrate this in any number of ways (depending on the form of externalism in question). Take a hard-line reliabilist who thinks that if you have no counterevidence for p, having an available reliable process for p is enough to justify you in believing p -- the more reliable that process is the more justified p is for you, and the more epistemically probable p is for you. Imagine Norman is offered a bet on the proposition that the President is in NY. For this externalist, since there is high reliability, the epistemic probability is high. And we can imagine that, in light of this high probability, accepting the bet has a higher expected value than refusing it. But now take a counterpart of Norman, Twin Norman, who is internally just like Norman but whose "clairvoyance" is not at all reliable. Well, this guy will have a low degree of justification (according to our externalist) and assuming the bet is the same and the stakes the same, he will have a higher expected value for refusing the bet. (Side note. Could original Normal be justified in taking the bet on p, even though Twin Norman isn't? My intuition: that's pretty implausible.) The argument above (before the side note) requires several assumptions. First, it would require an assumption about justified action: you are justified in doing what has the highest expected value of your available options. Second, it would require an assumption that the probabilities involved in that expectation are epistemic probabilities. Third, it would require the assumption that if epistemic probabilities are grounded in externalistic facts we can vary them holding fixed the internal stuff. If you give me these assumptions, then I can make a case that moderate externalism about epistemic probability has the consequence that there can be internal duplicates with the same options available to them who differ in what they are justified in doing. Bottom line: if you go externalist about epistemic probability and epsitemic probability feeds into expected value and justified action is a matter of expected value, then it shouldn't be surprising if you get the result that internal duplicates can differ in what they are justified in doing. I think that's bad for this sort of externalism, and hasn't really been noticed in the literature. Thanks for pushing Fantl and me on our brief remarks about this in the book. The matter deserves a more extended discussion than we gave it there. Ok, so we do agree that, given JJ, fallibilism, and moderate externalism, we get the result Fantl and I mentioned (i.e., that there can be internal duplicates, with the same options available, who differ in what they are justified in doing). I outlined how we get this result in my comment #3 — but, I agree, that to fallibilism has to be in there. We could talk about that. I take that to be an assumption that a lot of externalists would want to make (you don’t need epistemic probability 1 = justified certainty) to have justified belief.

The more I think about it, though, I wonder if we even *need* to bring in JJ to get a problem for moderate externalism. Consider moderate externalism about epistemic probability (or degree of justification if you like). This would hold that how probable p was for you was a function of externalistic factors. Now, take someone for whom p has a .7 probability, and suppose given her stakes, the option of doing A has an higher expected value than the option of doing B (and suppose those are the only choices). Call this person S1. It seems to me it should be possible for there to be a subject S2 with the same stakes and available options and who is internally indistinguishable from S1 but whose probability is very different (say .1), because of differences in the reliability of the available processes, or degree of safety of the various indications available, etc. We can imagine that for S2 the option of doing B has a higher expected value than that of doing A because of this difference in the probability of p for them. Then S1 and S2 would be alike internally, have the same options available, but differ in what they are justified in doing.

You can illustrate this in any number of ways (depending on the form of externalism in question). Take a hard-line reliabilist who thinks that if you have no counterevidence for p, having an available reliable process for p is enough to justify you in believing p — the more reliable that process is the more justified p is for you, and the more epistemically probable p is for you. Imagine Norman is offered a bet on the proposition that the President is in NY. For this externalist, since there is high reliability, the epistemic probability is high. And we can imagine that, in light of this high probability, accepting the bet has a higher expected value than refusing it. But now take a counterpart of Norman, Twin Norman, who is internally just like Norman but whose “clairvoyance” is not at all reliable. Well, this guy will have a low degree of justification (according to our externalist) and assuming the bet is the same and the stakes the same, he will have a higher expected value for refusing the bet.

(Side note. Could original Normal be justified in taking the bet on p, even though Twin Norman isn’t? My intuition: that’s pretty implausible.)

The argument above (before the side note) requires several assumptions. First, it would require an assumption about justified action: you are justified in doing what has the highest expected value of your available options. Second, it would require an assumption that the probabilities involved in that expectation are epistemic probabilities. Third, it would require the assumption that if epistemic probabilities are grounded in externalistic facts we can vary them holding fixed the internal stuff. If you give me these assumptions, then I can make a case that moderate externalism about epistemic probability has the consequence that there can be internal duplicates with the same options available to them who differ in what they are justified in doing.

Bottom line: if you go externalist about epistemic probability and epsitemic probability feeds into expected value and justified action is a matter of expected value, then it shouldn’t be surprising if you get the result that internal duplicates can differ in what they are justified in doing. I think that’s bad for this sort of externalism, and hasn’t really been noticed in the literature.

Thanks for pushing Fantl and me on our brief remarks about this in the book. The matter deserves a more extended discussion than we gave it there.

]]> By: Jonathan http://jonathanichikawa.net/weblog/justification-and-action/#comment-284 Jonathan Fri, 12 Feb 2010 10:14:02 +0000 http://jonathanichikawa.net/?p=189#comment-284 Matt, I agree with all of comment 13, but I don't see why it's relevant. Remember, I'm arguing that the combination of moderate externalism and (JJ) is not, as you say it is, implausible. It's not part of my defense of externalism to assume fallibilism. As you say, if you're a fallibilist, then it looks like the high probability should suffice for justified belief. The argument of the book is that this idea is inconsistent with plausible principles along the lines of (JJ). I'm not denying that; I'm inclined to think that something along those lines is probably right. I just don't see that this has anything to do with externalism. Matt, I agree with all of comment 13, but I don’t see why it’s relevant. Remember, I’m arguing that the combination of moderate externalism and (JJ) is not, as you say it is, implausible. It’s not part of my defense of externalism to assume fallibilism. As you say, if you’re a fallibilist, then it looks like the high probability should suffice for justified belief. The argument of the book is that this idea is inconsistent with plausible principles along the lines of (JJ). I’m not denying that; I’m inclined to think that something along those lines is probably right. I just don’t see that this has anything to do with externalism.

]]> By: Matt McGrath http://jonathanichikawa.net/weblog/justification-and-action/#comment-282 Matt McGrath Thu, 11 Feb 2010 20:52:33 +0000 http://jonathanichikawa.net/?p=189#comment-282 The thing is that in High 2 in order to have the expected value reason you'll need really high probability for p. Here's the expected value reason: EV(waiting till tomorrow) > EV(standing in line today). This boils down to the claim that the first sum below is greater than the second: Pr(it's open tomorrow)V(waiting till tomorrow when it's open tomorrow) + Pr(it's not open tomorrow)V(waiting till tomorrow when it's not open tomorrow) Pr(it's open tomorrow)V(standing in line today when it's open tomorrow) + Pr(it's not open tomorrow)V(standing in line today when it isn't open tomorrow). The only way to know or have good reason to think the first sum is greater than the second, in a standard High 2 kind of case, is to know or have good reason to think Pr(it's open tomorrow) is very very high -- surely high enough for justified belief by most fallibilist standards. Why is this? Because V(waiting till tomorrow when it isn't open tomorrow) is very very low, whereas V(standing in line today when it is open tomorrow) isn't all that low. So the probability for *it isn't open tomorrow* needs to be tiny, and so the probability for *it's open tomorrow* needs to be high. The thing is that in High 2 in order to have the expected value reason you’ll need really high probability for p. Here’s the expected value reason:

EV(waiting till tomorrow) > EV(standing in line today).

This boils down to the claim that the first sum below is greater than the second:

Pr(it’s open tomorrow)V(waiting till tomorrow when it’s open tomorrow) + Pr(it’s not open tomorrow)V(waiting till tomorrow when it’s not open tomorrow)

Pr(it’s open tomorrow)V(standing in line today when it’s open tomorrow) + Pr(it’s not open tomorrow)V(standing in line today when it isn’t open tomorrow).

The only way to know or have good reason to think the first sum is greater than the second, in a standard High 2 kind of case, is to know or have good reason to think Pr(it’s open tomorrow) is very very high — surely high enough for justified belief by most fallibilist standards.

Why is this? Because V(waiting till tomorrow when it isn’t open tomorrow) is very very low, whereas V(standing in line today when it is open tomorrow) isn’t all that low. So the probability for *it isn’t open tomorrow* needs to be tiny, and so the probability for *it’s open tomorrow* needs to be high.

]]> By: Jonathan http://jonathanichikawa.net/weblog/justification-and-action/#comment-280 Jonathan Wed, 10 Feb 2010 11:28:52 +0000 http://jonathanichikawa.net/?p=189#comment-280 Matt, I think there's room to deny that in High 2, one need be justified in believing that the bank will be open in order to justifiably wait until tomorrow. Suppose I justifiably believe that given my practical interests and time constraints, the expected value of waiting until tomorrow is positive. That looks like enough to me. Matt, I think there’s room to deny that in High 2, one need be justified in believing that the bank will be open in order to justifiably wait until tomorrow. Suppose I justifiably believe that given my practical interests and time constraints, the expected value of waiting until tomorrow is positive. That looks like enough to me.

]]> By: Jonathan http://jonathanichikawa.net/weblog/justification-and-action/#comment-279 Jonathan Wed, 10 Feb 2010 11:26:46 +0000 http://jonathanichikawa.net/?p=189#comment-279 Jeremy, I really don't see the issue here. What it is to be a moderate externalist is to think that not everything relevant to justification is internal. There's no tension at all in the idea that some things relevant to justification are internal. Jeremy, I really don’t see the issue here. What it is to be a moderate externalist is to think that not everything relevant to justification is internal. There’s no tension at all in the idea that some things relevant to justification are internal.

]]>